notebooks/word_stress_pytorch.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import time\n",
    "import glob\n",
    "import numba\n",
    "import random\n",
    "import itertools\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from tqdm import tqdm\n",
    "from sklearn.model_selection import train_test_split\n",
    "from torch.nn.utils.rnn import pad_sequence\n",
    "from torch.utils.data import Dataset, DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'1.11.0+cu113'"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_PATH = '../model/'\n",
    "RANDOM_SEED = 42"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def seed_everything(seed=0):\n",
    "    random.seed(seed)\n",
    "    os.environ['PYTHONHASHSEED'] = str(seed)\n",
    "    np.random.seed(seed)\n",
    "    torch.manual_seed(seed)\n",
    "    torch.cuda.manual_seed(seed)\n",
    "    torch.backends.cudnn.deterministic = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "seed_everything(RANDOM_SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# @numba.jit\n",
    "def stress_pos(x):\n",
    "    try:\n",
    "        res = np.zeros(len(x)-1)\n",
    "        stress_idx = x.find('+')\n",
    "        res[stress_idx] = 1\n",
    "        return res\n",
    "    except IndexError:\n",
    "        return np.NaN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def replace_accents(x):\n",
    "    chars = list(x.encode('utf-8').replace(b'\\xcc\\x81', b'+').decode('utf-8'))\n",
    "\n",
    "    final_chars = []\n",
    "    for i, c in enumerate(chars):\n",
    "        if c == '+':\n",
    "            try:\n",
    "                tmp = final_chars[i - 1]\n",
    "\n",
    "                final_chars.pop()\n",
    "                final_chars.append('+')\n",
    "                final_chars.append(tmp)\n",
    "            except IndexError:\n",
    "                final_chars.append(c)\n",
    "        else:\n",
    "            final_chars.append(c)\n",
    "\n",
    "    return ''.join(final_chars)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1/1 [00:09<00:00,  9.58s/it]\n"
     ]
    }
   ],
   "source": [
    "data = []\n",
    "\n",
    "for file in tqdm(glob.glob(DATA_PATH + \"*.txt\")):\n",
    "    with open(file) as fil:\n",
    "        lines = fil.readlines()\n",
    "        for line in lines:\n",
    "            try:\n",
    "                word2 = line.strip()\n",
    "                word2 = replace_accents(word2)\n",
    "                word1 = word2.replace(\"+\",\"\")\n",
    "                word1, word2 = word1.strip().replace('i', 'і'), word2.strip().replace('i', 'і')\n",
    "                data.append((word1, word2))\n",
    "            except ValueError:\n",
    "                continue"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.DataFrame(data)\n",
    "data.columns = ['word', 'word_with_stress']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 53/53 [00:54<00:00,  1.03s/it]\n"
     ]
    }
   ],
   "source": [
    "exclude_list = [\n",
    "    '\\t', '$', '%', '-', '\\\\', '`', 'a', 'b', 'c', 'd', 'e', 'f', 'g', \n",
    "    'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', \n",
    "    'v', 'w', 'x', 'y', 'z', '~', '\\xa0', '\\xad', '·', 'ђ', 'ј', 'љ', 'њ', 'ћ', \n",
    "    'ў', 'џ', '–', '‘', '’', '“', '”', '•', '№',\n",
    "    'ъ', 'ы', 'э'\n",
    "]\n",
    "\n",
    "for item in tqdm(exclude_list):\n",
    "    data = data.loc[~data.word.str.contains(item, regex=False)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 6.42 s, sys: 492 ms, total: 6.91 s\n",
      "Wall time: 6.89 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "data['word_list'] = data.word.map(list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 3.06 s, sys: 63.6 ms, total: 3.13 s\n",
      "Wall time: 3.13 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "data['word_stress_pos'] = data.word_with_stress.map(stress_pos)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.loc[data.word_stress_pos.notna()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>word</th>\n",
       "      <th>word_with_stress</th>\n",
       "      <th>word_list</th>\n",
       "      <th>word_stress_pos</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2906284</th>\n",
       "      <td>шкитильгатимете</td>\n",
       "      <td>шкитильг+атимете</td>\n",
       "      <td>[ш, к, и, т, и, л, ь, г, а, т, и, м, е, т, е]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1285208</th>\n",
       "      <td>надгризи</td>\n",
       "      <td>надгр+изи</td>\n",
       "      <td>[н, а, д, г, р, и, з, и]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>219867</th>\n",
       "      <td>вгнімось</td>\n",
       "      <td>вгн+імось</td>\n",
       "      <td>[в, г, н, і, м, о, с, ь]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>648244</th>\n",
       "      <td>достарчиш</td>\n",
       "      <td>дост+арчиш</td>\n",
       "      <td>[д, о, с, т, а, р, ч, и, ш]</td>\n",
       "      <td>[0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2505282</th>\n",
       "      <td>сталінку</td>\n",
       "      <td>ст+алінку</td>\n",
       "      <td>[с, т, а, л, і, н, к, у]</td>\n",
       "      <td>[0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    word  word_with_stress  \\\n",
       "2906284  шкитильгатимете  шкитильг+атимете   \n",
       "1285208         надгризи         надгр+изи   \n",
       "219867          вгнімось         вгн+імось   \n",
       "648244         достарчиш        дост+арчиш   \n",
       "2505282         сталінку         ст+алінку   \n",
       "\n",
       "                                             word_list  \\\n",
       "2906284  [ш, к, и, т, и, л, ь, г, а, т, и, м, е, т, е]   \n",
       "1285208                       [н, а, д, г, р, и, з, и]   \n",
       "219867                        [в, г, н, і, м, о, с, ь]   \n",
       "648244                     [д, о, с, т, а, р, ч, и, ш]   \n",
       "2505282                       [с, т, а, л, і, н, к, у]   \n",
       "\n",
       "                                           word_stress_pos  \n",
       "2906284  [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, ...  \n",
       "1285208           [0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0]  \n",
       "219867            [0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0]  \n",
       "648244       [0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0]  \n",
       "2505282           [0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0]  "
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "29"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "max_sequence_len = np.max(data.word.str.len())\n",
    "max_sequence_len"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def flatten(array):\n",
    "    for item in array:\n",
    "        if isinstance(item, list):\n",
    "            yield from flatten(item)\n",
    "        else:\n",
    "            yield item\n",
    "\n",
    "\n",
    "class SequenceTokenizer:\n",
    "    \n",
    "    def __init__(self):\n",
    "        self.word2index = {}\n",
    "        self.index2word = {}\n",
    "        self.oov_token ='<UNK>'\n",
    "        self.oov_token_index = 0\n",
    "        \n",
    "    def fit(self, sequence):\n",
    "        self.index2word = dict(enumerate([self.oov_token] + sorted(set(flatten(sequence))), 1))\n",
    "        self.word2index = {v:k for k,v in self.index2word.items()}\n",
    "        self.oov_token_index = self.word2index.get(self.oov_token)\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X):\n",
    "        res = []\n",
    "        for line in X:\n",
    "            res.append([self.word2index.get(item, self.oov_token_index) for item in line])\n",
    "        return res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = SequenceTokenizer()\n",
    "tokenizer.fit(data.word_list);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'<UNK>': 1,\n",
       " \"'\": 2,\n",
       " 'а': 3,\n",
       " 'б': 4,\n",
       " 'в': 5,\n",
       " 'г': 6,\n",
       " 'д': 7,\n",
       " 'е': 8,\n",
       " 'ж': 9,\n",
       " 'з': 10,\n",
       " 'и': 11,\n",
       " 'й': 12,\n",
       " 'к': 13,\n",
       " 'л': 14,\n",
       " 'м': 15,\n",
       " 'н': 16,\n",
       " 'о': 17,\n",
       " 'п': 18,\n",
       " 'р': 19,\n",
       " 'с': 20,\n",
       " 'т': 21,\n",
       " 'у': 22,\n",
       " 'ф': 23,\n",
       " 'х': 24,\n",
       " 'ц': 25,\n",
       " 'ч': 26,\n",
       " 'ш': 27,\n",
       " 'щ': 28,\n",
       " 'ь': 29,\n",
       " 'ю': 30,\n",
       " 'я': 31,\n",
       " 'є': 32,\n",
       " 'і': 33,\n",
       " 'ї': 34,\n",
       " 'ґ': 35}"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenizer.word2index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = tokenizer.transform(data.word_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def pad_sequence(lst, max_seq=max_sequence_len):\n",
    "    if isinstance(lst[0], list):\n",
    "        return np.array([i + [0]*(max_seq-len(i)) for i in lst])\n",
    "    else:\n",
    "        lst + [0]*(max_seq-len(lst))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %%time\n",
    "\n",
    "# input_seq = zip(*itertools.zip_longest(*X, fillvalue=0))\n",
    "# input_seq = list(map(list, input_seq))\n",
    "# input_seq = np.array(input_seq)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 10.9 s, sys: 340 ms, total: 11.2 s\n",
      "Wall time: 11.2 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "input_seq = pad_sequence(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2960052, 29)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[ 3,  0,  0, ...,  0,  0,  0],\n",
       "       [ 3,  3, 14, ...,  0,  0,  0],\n",
       "       [ 3,  3, 14, ...,  0,  0,  0],\n",
       "       ...,\n",
       "       [35, 22, 19, ...,  0,  0,  0],\n",
       "       [35, 22, 19, ...,  0,  0,  0],\n",
       "       [35, 22, 19, ...,  0,  0,  0]])"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(input_seq.shape)\n",
    "input_seq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = data.word_stress_pos.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "output_seq = zip(*itertools.zip_longest(*y, fillvalue=0))\n",
    "output_seq = list(map(list, output_seq))\n",
    "output_seq = np.array(output_seq).astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2960052, 29)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[1, 0, 0, ..., 0, 0, 0],\n",
       "       [1, 0, 0, ..., 0, 0, 0],\n",
       "       [1, 0, 0, ..., 0, 0, 0],\n",
       "       ...,\n",
       "       [0, 0, 0, ..., 0, 0, 0],\n",
       "       [0, 0, 0, ..., 0, 0, 0],\n",
       "       [0, 0, 0, ..., 0, 0, 0]])"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(output_seq.shape)\n",
    "output_seq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "(input_seq_train, input_seq_val, \n",
    " output_seq_train, output_seq_val) = train_test_split(input_seq, \n",
    "                                                      output_seq, \n",
    "                                                      test_size=0.5, \n",
    "                                                      random_state=RANDOM_SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "input_seq_train = torch.tensor(input_seq_train, dtype=torch.long).cuda()\n",
    "input_seq_val = torch.tensor(input_seq_val, dtype=torch.long).cuda()\n",
    "output_seq_train = torch.tensor(output_seq_train, dtype=torch.float).cuda()\n",
    "output_seq_val = torch.tensor(output_seq_val, dtype=torch.float).cuda()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "class MyDataset(Dataset):\n",
    "    def __init__(self, dataset):\n",
    "        self.dataset = dataset\n",
    "        \n",
    "    def __getitem__(self, index):\n",
    "        data,target = self.dataset[index]\n",
    "        return data, target, index\n",
    "    \n",
    "    def __len__(self):\n",
    "        return len(self.dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "class LSTM_model(nn.Module):\n",
    "\n",
    "    def __init__(self, embedding_dim, hidden_dim, vocab_size, target_size):\n",
    "        super(LSTM_model, self).__init__()\n",
    "        self.hidden_dim = hidden_dim\n",
    "\n",
    "        self.embeddings = nn.Embedding(vocab_size, embedding_dim)\n",
    "\n",
    "        self.lstm = nn.LSTM(input_size=self.embeddings.embedding_dim,\n",
    "                            hidden_size=hidden_dim,\n",
    "                            num_layers=3,\n",
    "                            batch_first=True,\n",
    "                            bidirectional=True,\n",
    "                            dropout = 0.05)\n",
    "        self.linear = nn.Linear(self.hidden_dim * 8 , 64)\n",
    "        self.batch_norm = nn.BatchNorm1d(self.hidden_dim * 8, affine=False)\n",
    "        self.relu = nn.ReLU()\n",
    "        self.dropout = nn.Dropout(0.1)\n",
    "        self.out = nn.Linear(64, target_size)\n",
    "\n",
    "    def forward(self, x):\n",
    "        h_embeddings = self.embeddings(x)\n",
    "        \n",
    "        h_lstm, _ = self.lstm(h_embeddings)\n",
    "        d_1 = h_lstm[:,0,:]\n",
    "        d_2 = h_lstm[:,h_lstm.shape[1]//4,:]\n",
    "        d_3 = h_lstm[:,h_lstm.shape[1]*3//4,:]\n",
    "        d_4 = h_lstm[:,-1,:]\n",
    "        x = torch.cat((d_1, d_2, d_3, d_4), 1)\n",
    "        x = self.batch_norm(x)\n",
    "        x = self.linear(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.dropout(x)\n",
    "        x = self.out(x)\n",
    "        y = nn.functional.softmax(x, dim=1)\n",
    "        return y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LSTM_model(\n",
       "  (embeddings): Embedding(36, 64)\n",
       "  (lstm): LSTM(64, 64, num_layers=3, batch_first=True, dropout=0.05, bidirectional=True)\n",
       "  (linear): Linear(in_features=512, out_features=64, bias=True)\n",
       "  (batch_norm): BatchNorm1d(512, eps=1e-05, momentum=0.1, affine=False, track_running_stats=True)\n",
       "  (relu): ReLU()\n",
       "  (dropout): Dropout(p=0.1, inplace=False)\n",
       "  (out): Linear(in_features=64, out_features=29, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = LSTM_model(embedding_dim=64, \n",
    "                   hidden_dim=64, \n",
    "                   vocab_size=len(tokenizer.word2index) + 1, \n",
    "                   target_size=max_sequence_len)\n",
    "model.cuda()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_function = nn.BCEWithLogitsLoss()\n",
    "optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()), lr=1e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "BATCH_SIZE = 256 * 2\n",
    "\n",
    "train = MyDataset(torch.utils.data.TensorDataset(input_seq_train, output_seq_train))\n",
    "valid = MyDataset(torch.utils.data.TensorDataset(input_seq_val, output_seq_val))\n",
    "\n",
    "train_loader = torch.utils.data.DataLoader(train, batch_size=BATCH_SIZE, shuffle=True)\n",
    "valid_loader = torch.utils.data.DataLoader(valid, batch_size=BATCH_SIZE, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# with torch.no_grad():\n",
    "#     (x_sample, y_sample, index) = train[:2]\n",
    "#     preds = model(x_sample.reshape(1, -1))\n",
    "#     print(preds.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### K-Fold training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import KFold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_folds = 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "kf = KFold(n_splits=n_folds, shuffle=True, random_state=RANDOM_SEED)\n",
    "# kf.get_n_splits(X=input_seq, y=output_seq)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2368041,) (592011,)\n",
      "(2368041,) (592011,)\n",
      "(2368042,) (592010,)\n",
      "(2368042,) (592010,)\n",
      "(2368042,) (592010,)\n"
     ]
    }
   ],
   "source": [
    "for train_index, test_index in kf.split(X=input_seq, y=output_seq):\n",
    "    print(train_index.shape, test_index.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2960052, 29)"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "input_seq"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch [1/50]  progress = 100% \t loss=0.6887 \t acc=74.64% \n",
      "Epoch [1/50]  results:\t\t loss=0.6887\t acc=74.64%\t val_loss=0.6867\t val_acc=80.76%\t time=93.47s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [2/50]  progress = 100% \t loss=0.6862 \t acc=82.03% \n",
      "Epoch [2/50]  results:\t\t loss=0.6862\t acc=82.03%\t val_loss=0.6859\t val_acc=83.07%\t time=94.58s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [3/50]  progress = 100% \t loss=0.6857 \t acc=83.68% \n",
      "Epoch [3/50]  results:\t\t loss=0.6857\t acc=83.68%\t val_loss=0.6854\t val_acc=84.42%\t time=95.14s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [4/50]  progress = 100% \t loss=0.6853 \t acc=84.74% \n",
      "Epoch [4/50]  results:\t\t loss=0.6853\t acc=84.74%\t val_loss=0.6851\t val_acc=85.22%\t time=94.73s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [5/50]  progress = 100% \t loss=0.6850 \t acc=85.49% \n",
      "Epoch [5/50]  results:\t\t loss=0.6850\t acc=85.49%\t val_loss=0.6849\t val_acc=85.89%\t time=94.77s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [6/50]  progress = 100% \t loss=0.6848 \t acc=86.15% \n",
      "Epoch [6/50]  results:\t\t loss=0.6848\t acc=86.15%\t val_loss=0.6847\t val_acc=86.34%\t time=94.42s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [7/50]  progress = 100% \t loss=0.6847 \t acc=86.56% \n",
      "Epoch [7/50]  results:\t\t loss=0.6847\t acc=86.56%\t val_loss=0.6846\t val_acc=86.73%\t time=93.97s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [8/50]  progress = 100% \t loss=0.6845 \t acc=86.95% \n",
      "Epoch [8/50]  results:\t\t loss=0.6845\t acc=86.95%\t val_loss=0.6846\t val_acc=86.89%\t time=92.12s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [9/50]  progress = 100% \t loss=0.6844 \t acc=87.32% \n",
      "Epoch [9/50]  results:\t\t loss=0.6844\t acc=87.32%\t val_loss=0.6845\t val_acc=87.16%\t time=94.89s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [10/50]  progress = 100% \t loss=0.6843 \t acc=87.59% \n",
      "Epoch [10/50]  results:\t\t loss=0.6843\t acc=87.59%\t val_loss=0.6842\t val_acc=87.88%\t time=94.04s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [11/50]  progress = 100% \t loss=0.6842 \t acc=87.90% \n",
      "Epoch [11/50]  results:\t\t loss=0.6842\t acc=87.90%\t val_loss=0.6841\t val_acc=88.13%\t time=93.60s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [12/50]  progress = 100% \t loss=0.6841 \t acc=88.19% \n",
      "Epoch [12/50]  results:\t\t loss=0.6841\t acc=88.19%\t val_loss=0.6841\t val_acc=88.18%\t time=95.54s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [13/50]  progress = 100% \t loss=0.6841 \t acc=88.26% \n",
      "Epoch [13/50]  results:\t\t loss=0.6841\t acc=88.26%\t val_loss=0.6840\t val_acc=88.40%\t time=95.17s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [14/50]  progress = 100% \t loss=0.6840 \t acc=88.53% \n",
      "Epoch [14/50]  results:\t\t loss=0.6840\t acc=88.53%\t val_loss=0.6840\t val_acc=88.46%\t time=95.14s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [15/50]  progress = 100% \t loss=0.6839 \t acc=88.66% \n",
      "Epoch [15/50]  results:\t\t loss=0.6839\t acc=88.66%\t val_loss=0.6839\t val_acc=88.82%\t time=95.14s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [16/50]  progress = 100% \t loss=0.6839 \t acc=88.83% \n",
      "Epoch [16/50]  results:\t\t loss=0.6839\t acc=88.83%\t val_loss=0.6839\t val_acc=88.74%\t time=95.38s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [17/50]  progress = 100% \t loss=0.6839 \t acc=88.83% \n",
      "Epoch [17/50]  results:\t\t loss=0.6839\t acc=88.83%\t val_loss=0.6839\t val_acc=88.91%\t time=94.70s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [18/50]  progress = 100% \t loss=0.6838 \t acc=89.03% \n",
      "Epoch [18/50]  results:\t\t loss=0.6838\t acc=89.03%\t val_loss=0.6838\t val_acc=89.12%\t time=95.82s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [19/50]  progress = 100% \t loss=0.6838 \t acc=89.06% \n",
      "Epoch [19/50]  results:\t\t loss=0.6838\t acc=89.06%\t val_loss=0.6838\t val_acc=89.09%\t time=94.26s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [20/50]  progress = 100% \t loss=0.6838 \t acc=89.09% \n",
      "Epoch [20/50]  results:\t\t loss=0.6838\t acc=89.09%\t val_loss=0.6838\t val_acc=89.19%\t time=95.72s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [21/50]  progress = 100% \t loss=0.6838 \t acc=89.20% \n",
      "Epoch [21/50]  results:\t\t loss=0.6838\t acc=89.20%\t val_loss=0.6838\t val_acc=89.16%\t time=94.04s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [22/50]  progress = 100% \t loss=0.6837 \t acc=89.25% \n",
      "Epoch [22/50]  results:\t\t loss=0.6837\t acc=89.25%\t val_loss=0.6838\t val_acc=89.23%\t time=95.10s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [23/50]  progress = 100% \t loss=0.6837 \t acc=89.30% \n",
      "Epoch [23/50]  results:\t\t loss=0.6837\t acc=89.30%\t val_loss=0.6837\t val_acc=89.47%\t time=95.06s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [24/50]  progress = 100% \t loss=0.6837 \t acc=89.32% \n",
      "Epoch [24/50]  results:\t\t loss=0.6837\t acc=89.32%\t val_loss=0.6837\t val_acc=89.44%\t time=92.15s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [25/50]  progress = 100% \t loss=0.6837 \t acc=89.47% \n",
      "Epoch [25/50]  results:\t\t loss=0.6837\t acc=89.47%\t val_loss=0.6837\t val_acc=89.44%\t time=94.11s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [26/50]  progress = 100% \t loss=0.6837 \t acc=89.43% \n",
      "Epoch [26/50]  results:\t\t loss=0.6837\t acc=89.43%\t val_loss=0.6836\t val_acc=89.58%\t time=94.53s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [27/50]  progress = 100% \t loss=0.6836 \t acc=89.57% \n",
      "Epoch [27/50]  results:\t\t loss=0.6836\t acc=89.57%\t val_loss=0.6837\t val_acc=89.53%\t time=95.77s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [28/50]  progress = 100% \t loss=0.6837 \t acc=89.52% \n",
      "Epoch [28/50]  results:\t\t loss=0.6837\t acc=89.52%\t val_loss=0.6836\t val_acc=89.56%\t time=94.99s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [29/50]  progress = 100% \t loss=0.6836 \t acc=89.60% \n",
      "Epoch [29/50]  results:\t\t loss=0.6836\t acc=89.60%\t val_loss=0.6836\t val_acc=89.62%\t time=93.64s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [30/50]  progress = 100% \t loss=0.6837 \t acc=89.53% \n",
      "Epoch [30/50]  results:\t\t loss=0.6837\t acc=89.53%\t val_loss=0.6837\t val_acc=89.51%\t time=96.26s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [31/50]  progress = 100% \t loss=0.6837 \t acc=89.52% \n",
      "Epoch [31/50]  results:\t\t loss=0.6837\t acc=89.52%\t val_loss=0.6836\t val_acc=89.58%\t time=95.65s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [32/50]  progress = 100% \t loss=0.6836 \t acc=89.55% \n",
      "Epoch [32/50]  results:\t\t loss=0.6836\t acc=89.55%\t val_loss=0.6837\t val_acc=89.42%\t time=94.99s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [33/50]  progress = 100% \t loss=0.6837 \t acc=89.46% \n",
      "Epoch [33/50]  results:\t\t loss=0.6837\t acc=89.46%\t val_loss=0.6837\t val_acc=89.49%\t time=93.73s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [34/50]  progress = 100% \t loss=0.6837 \t acc=89.48% \n",
      "Epoch [34/50]  results:\t\t loss=0.6837\t acc=89.48%\t val_loss=0.6837\t val_acc=89.53%\t time=95.17s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [35/50]  progress = 100% \t loss=0.6837 \t acc=89.52% \n",
      "Epoch [35/50]  results:\t\t loss=0.6837\t acc=89.52%\t val_loss=0.6836\t val_acc=89.64%\t time=95.05s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [36/50]  progress = 100% \t loss=0.6836 \t acc=89.62% \n",
      "Epoch [36/50]  results:\t\t loss=0.6836\t acc=89.62%\t val_loss=0.6837\t val_acc=89.48%\t time=94.62s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [37/50]  progress = 100% \t loss=0.6836 \t acc=89.54% \n",
      "Epoch [37/50]  results:\t\t loss=0.6836\t acc=89.54%\t val_loss=0.6836\t val_acc=89.58%\t time=93.38s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [38/50]  progress = 100% \t loss=0.6836 \t acc=89.58% \n",
      "Epoch [38/50]  results:\t\t loss=0.6836\t acc=89.58%\t val_loss=0.6837\t val_acc=89.50%\t time=92.91s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [39/50]  progress = 100% \t loss=0.6837 \t acc=89.46% \n",
      "Epoch [39/50]  results:\t\t loss=0.6837\t acc=89.46%\t val_loss=0.6837\t val_acc=89.52%\t time=92.17s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [40/50]  progress = 100% \t loss=0.6837 \t acc=89.36% \n",
      "Epoch [40/50]  results:\t\t loss=0.6837\t acc=89.36%\t val_loss=0.6837\t val_acc=89.34%\t time=94.65s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [41/50]  progress = 100% \t loss=0.6837 \t acc=89.46% \n",
      "Epoch [41/50]  results:\t\t loss=0.6837\t acc=89.46%\t val_loss=0.6837\t val_acc=89.31%\t time=94.26s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [42/50]  progress = 100% \t loss=0.6837 \t acc=89.39% \n",
      "Epoch [42/50]  results:\t\t loss=0.6837\t acc=89.39%\t val_loss=0.6838\t val_acc=89.22%\t time=93.31s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [43/50]  progress = 100% \t loss=0.6837 \t acc=89.29% \n",
      "Epoch [43/50]  results:\t\t loss=0.6837\t acc=89.29%\t val_loss=0.6837\t val_acc=89.53%\t time=93.99s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [44/50]  progress = 100% \t loss=0.6837 \t acc=89.38% \n",
      "Epoch [44/50]  results:\t\t loss=0.6837\t acc=89.38%\t val_loss=0.6837\t val_acc=89.29%\t time=95.62s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [45/50]  progress = 100% \t loss=0.6837 \t acc=89.33% \n",
      "Epoch [45/50]  results:\t\t loss=0.6837\t acc=89.33%\t val_loss=0.6837\t val_acc=89.43%\t time=95.91s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [46/50]  progress = 100% \t loss=0.6837 \t acc=89.31% \n",
      "Epoch [46/50]  results:\t\t loss=0.6837\t acc=89.31%\t val_loss=0.6838\t val_acc=89.15%\t time=95.05s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [47/50]  progress = 100% \t loss=0.6838 \t acc=89.19% \n",
      "Epoch [47/50]  results:\t\t loss=0.6838\t acc=89.19%\t val_loss=0.6837\t val_acc=89.26%\t time=97.26s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [48/50]  progress = 100% \t loss=0.6837 \t acc=89.27% \n",
      "Epoch [48/50]  results:\t\t loss=0.6837\t acc=89.27%\t val_loss=0.6837\t val_acc=89.27%\t time=97.17s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [49/50]  progress = 100% \t loss=0.6838 \t acc=89.22% \n",
      "Epoch [49/50]  results:\t\t loss=0.6838\t acc=89.22%\t val_loss=0.6837\t val_acc=89.33%\t time=94.92s\n",
      "------------------------------------------------------------------------------\n",
      "Epoch [50/50]  progress = 100% \t loss=0.6837 \t acc=89.26% \n",
      "Epoch [50/50]  results:\t\t loss=0.6837\t acc=89.26%\t val_loss=0.6837\t val_acc=89.43%\t time=95.63s\n",
      "------------------------------------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "n_epochs = 50\n",
    "history = {'train': {}, 'val': {}}\n",
    "teacher_forcing_ratio = 0.5\n",
    "\n",
    "for epoch in range(1, n_epochs + 1):\n",
    "    start_time = time.time()\n",
    "    \n",
    "    model.train()\n",
    "\n",
    "    avg_loss, total_loss, avg_acc, total_acc = 0., 0., 0., 0.\n",
    "    for i, (x_batch, y_batch, index) in enumerate(train_loader):\n",
    "        y_pred = model(x_batch)\n",
    "        loss = loss_function(y_pred, y_batch)\n",
    "        \n",
    "        optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        \n",
    "        equal = torch.eq(torch.argmax(y_pred, axis=1), torch.argmax(y_batch, axis=1))\n",
    "        batch_acc = int(equal.sum(-1)) / y_batch.shape[0]\n",
    "        batch_loss = loss.item()\n",
    "        \n",
    "        total_acc += batch_acc\n",
    "        total_loss += batch_loss\n",
    "        print(f\"\\rEpoch [{epoch}/{n_epochs}] \"\n",
    "              f\" progress = {round(i/len(train_loader)*100)}% \"\n",
    "              f\"\\t loss={total_loss / (i + 1):.4f} \"\n",
    "              f\"\\t acc={total_acc / (i + 1) * 100:.2f}% \", end='')\n",
    "    avg_loss = total_loss / len(train_loader)\n",
    "    avg_acc = total_acc / len(train_loader)\n",
    "    history['train']['loss'] = history.get('train', {}).get('loss', []) + [avg_loss]\n",
    "    history['train']['accuracy'] = history.get('train', {}).get('accuracy', []) + [avg_acc]\n",
    "        \n",
    "    model.eval()\n",
    "\n",
    "    \n",
    "    avg_val_loss, total_val_loss, avg_val_acc, total_val_acc = 0., 0., 0., 0.\n",
    "    for i, (x_batch, y_batch, index) in enumerate(valid_loader):\n",
    "        y_pred = model(x_batch).detach()\n",
    "        val_loss = loss_function(y_pred, y_batch)\n",
    "        \n",
    "        equal = torch.eq(torch.argmax(y_pred, axis=1), torch.argmax(y_batch, axis=1))\n",
    "        batch_val_acc = int(equal.sum(-1)) / y_batch.shape[0]\n",
    "        batch_val_loss = val_loss.item()\n",
    "        \n",
    "        total_val_acc += batch_val_acc\n",
    "        total_val_loss += batch_val_loss\n",
    "    avg_val_loss = total_val_loss / len(valid_loader)\n",
    "    avg_val_acc = total_val_acc / len(valid_loader)\n",
    "    history['val']['loss'] = history.get('val', {}).get('loss', []) + [avg_val_loss]\n",
    "    history['val']['accuracy'] = history.get('val', {}).get('accuracy', []) + [avg_val_acc]\n",
    "    \n",
    "    elapsed_time = time.time() - start_time \n",
    "    print(f\"\\nEpoch [{epoch}/{n_epochs}]  results:\"\n",
    "          f\"\\t\\t loss={avg_loss:.4f}\"\n",
    "          f\"\\t acc={avg_acc * 100:.2f}%\"\n",
    "          f\"\\t val_loss={avg_val_loss:.4f}\"\n",
    "          f\"\\t val_acc={avg_val_acc * 100:.2f}%\"\n",
    "          f\"\\t time={elapsed_time:.2f}s\")\n",
    "    print(\"-\"*78)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12, 8))\n",
    "\n",
    "plt.plot(history['train']['loss'])\n",
    "plt.plot(history['val']['loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('val')\n",
    "plt.xlabel('epoch')\n",
    "plt.xticks(np.arange(len(history['train']['loss'])), np.arange(1, len(history['train']['loss']) + 1))\n",
    "plt.legend(['train', 'val'], loc='upper right')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12, 8))\n",
    "\n",
    "plt.plot(history['train']['accuracy'])\n",
    "plt.plot(history['val']['accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.xticks(np.arange(len(history['train']['accuracy'])), np.arange(1, len(history['train']['accuracy']) + 1))\n",
    "plt.legend(['train', 'val'], loc='lower right')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(words, mode='stress'):\n",
    "    tokens = pad_sequence(tokenizer.transform(words))\n",
    "    sequences = torch.tensor(tokens, dtype=torch.long).cuda()\n",
    "    preds = model(sequences)\n",
    "    indeces = torch.argmax(preds, axis=1)\n",
    "    if mode == 'stress':\n",
    "        return [word[:index+1] + chr(769) + word[index+1:] for word, index in zip(words, indeces)]\n",
    "    elif mode == 'asterisk':\n",
    "        return [word[:index+1] + \"*\" + word[index+1:] for word, index in zip(words, indeces)]\n",
    "    else:\n",
    "        raise ValueError(f\"Wrong `mode`={mode}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "words = [\"словотворення\", \"архаїчний\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['словотво́рення', 'архаї́чний']"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predict(words)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.save(model.state_dict(),\"accentor.pt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.8.10 64-bit",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  },
  "vscode": {
   "interpreter": {
    "hash": "767d51c1340bd893661ea55ea3124f6de3c7a262a8b4abca0554b478b1e2ff90"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}